CUI ET AL . VOL. 6 ’ NO. 3 ’ 2385–2393 ’ 2012

www.acsnano.org

2385

February 16, 2012

C 2012 American Chemical Society

Dynamic Tuning and SymmetryLowering of Fano Resonance inPlasmonic NanostructureYonghao Cui, Jianhong Zhou, Venkata A. Tamma, and Wounjhang Park*

Department of Electrical, Computer and Energy Engineering, University of Colorado at Boulder, Boulder, Colorado 80309, United States

Complex metallic nanostructures sup-porting a collective surface plasmonresonance are currently a topic of

strong research interest. In this system, thesurface plasmon resonances of constituentelements couple together to form collectivemodes delocalized over the entire structure.The characteristics of the collective plasmonmodesare strong functionsof thedetails of thenanostructure, opening new opportunities toinvestigate the interaction between plasmonicnanostructures andalso toengineer thenatureof the resonance for various applications, forexample, in sensing and nonlinear devices.Much of the current research has been direc-ted tometal nanoparticle aggregates in whichthe nanoparticle size, spacing between thenanoparticles, and symmetry of the aggregatecould lead to different coupled plasmonmodes.1�14 The high sensitivity to the struc-tural parameters is of special interest becauseitmakes the system an excellent candidate formechanically tunable devices. Structural tun-ing by mechanical stress is a natural way toachieve wide tunability in artificial materialsand structures, which derive their propertiesfrom structural design. Mechanical tuning hasbeen applied to photonic crystals, which ex-hibit many novel properties stemming fromtheir periodicity.15,16 Recently, tunable nega-tive index imaging by a flexible photoniccrystal has been reported.17 Similar strategieshave also been applied to metamaterials inwhich plasmonic nanostructures were subjectto mechanical stress to distort and tune theirresonances and thereby achieve a tunablemetamaterial response.18,19 It therefore seemsonly natural to explore the mechanical tuningofplasmonicnanostructureswhose resonanceis highly sensitive to their structural para-meters, and mechanically tunable resonanceshave been demonstrated for a dolmen-typeresonator18 and nanoparticle dimer.20

Fano resonances,21 which arise from theinterference of a sharp resonance with a

broad background, have attracted a greatdeal of research interest recently due to theirasymmetric line shape, sharp resonance, andsensitivity to a variety of parameters. Tradi-tionally Fano resonances have been consid-ered mostly in quantum systems.22�24 How-ever, Fano resonances have recently beenrealized in plasmonic nanostructures andmetamaterials.2,5�14,18,25�31 Due to its narrowline width and high sensitivity to structuraland environmental parameters, a Fano reso-nance has great potential for photonic appli-cations suchas sensing.32 It hasbeen reportedthat in many nanoparticle aggregates thecollective plasmon modes exhibit a Fanoresonance.2,5�8,18 A heptamer is one of thereported metal nanoclusters that support aFano resonance,5�14 and it is predicted toexhibit a very large Fano resonance spectral

* Address correspondence towon.park@colorado.edu.

Received for review November 29, 2011and accepted February 16, 2012.

Published online10.1021/nn204647b

ABSTRACT We present dynamic tuning

and symmetry lowering of Fano resonances in

gold heptamers accomplished by applying

uniaxial mechanical stress. The flexible hep-

tamer structure was obtained by embedding

the seven-gold-nanocylinder complex in a polydimethylsiloxane membrane. Under uniaxial

stress, the Fano resonance exhibited opposite spectral shifts for the two orthogonal

polarizations parallel and perpendicular to the mechanical stress. Furthermore, a new

resonance was observed for polarization parallel to the mechanical stress but not for the

perpendicular polarization. The experimental results showed good agreement with the

numerical simulations. A detailed group theoretical analysis showed that the symmetry

lowering caused by the mechanical stress not only splits the originally degenerate mode but

also modifies the originally optically inactive mode into an optically active mode, which then

interacts strongly with a closely spaced mode and exhibits anticrossing behavior. The

symmetry tuning enabled by applying mechanical stress is a simple and efficient way to

engineer the nature of coupled plasmon resonances in complex nanostructures. The

mechanically tunable plasmonic nanostructures also provide an excellent platform for

dynamically tunable nanophotonic devices such as tunable filters and sensors.

KEYWORDS: surface plasmon . Fano resonance . mechanical tuning . plasmonicnanostructure . localized surface plasmon resonance . anticrossing

ARTIC

LE

CUI ET AL . VOL. 6 ’ NO. 3 ’ 2385–2393 ’ 2012

www.acsnano.org

2386

shift upon tuning of the surroundingmedia's refractiveindex.8 A Fano resonance in a heptamer is also highlysensitive to the gaps between themetal nanoparticles6

and the symmetry of the heptamer.8 Therefore, theheptamer structure is a good candidate to demon-strate a tunable Fano resonance through refractiveindex tuning, structural tuning, and symmetry tuning.For tuning efficiency, refractive index tuning is limitedby the small range of attainable refractive indexchanges, while mechanical tuning, which can tunethe gaps between metal nanoparticles and the sym-metry of the heptamer, can produce much widertunability and is capable of tuning structural para-meters and symmetry simultaneously. Symmetrybreaking is of particular interest, as it could not onlylead to resonance frequency shifts but also alter thefundamental characteristics of the modes, leading todramatic changes in the optical properties. Recentlythe symmetry-breaking effects have been studied forgold nanoparticle trimers.33,34

In this paper, we report a dynamically tunable Fanoresonance in a gold heptamer structure embedded in aflexible polymer membrane and present how thesymmetry lowering induced by mechanical stressaffects the tunability. The gold heptamer structure wascomposed of seven gold nanocylinders embedded in aflexible polydimethylsiloxane (PDMS) membrane.Thanks to the high elasticity of the PDMS membrane,the gaps between the gold nanocylinders could beaccurately tuned bymechanically stretching the PDMSmembrane, leading to tuning of the Fano resonance.The uniaxial stress also lowers the symmetry of theheptamer structures, causing the splitting of the orig-inally degenerate modes and also turning an originallyoptically inactive mode into an optically active mode.The resultant split modes and the newly createdoptically active modes strongly interact with oneanother and lead to distinct polarization depen-dence. In the following, we present experimentalresults on mechanical tuning of a Fano resonance ina gold heptamer and provide a detailed theoreticalstudy on the mode properties of the heptamerunder mechanical stress that explains the observedbehaviors.

RESULTS AND DISCUSSION

The gold heptamerwas designed to have seven goldnanocylinders with a diameter of 150 nm and height of80 nm. The gap between cylinders was designed to be30 nm. For reliable optical characterizations, a massivearray of gold heptamers was fabricated in an arraycovering an area of 400� 400 μm2. Themassive array isformed by an 8 � 8 array of heptamers with individualarray size of 50 � 50 μm2. Figure 1a shows a schematicdiagram of a free-standing PDMS membrane with anembedded array of heptamers. Scanning electronmicro-scopy (SEM) confirmed that high-quality heptamerswere

successfully fabricated with structural parametersclose to the design values. Figure 1b�d shows the SEMimages of the fabricated heptamer array. From the high-magnification SEM image shown in Figure 1c, the dia-meter of the gold cylinder was measured to be 147 nm.Figure 1c also shows that the heptamer structure wasshrunk by 3.7% along the vertical direction and the gapsbetween the cylinders were 33 nm along the horizontaldirection and 28 nm along the diagonal.The heptamers were subsequently stretched along

the horizontal direction in Figure 1c. The optical ex-tinction spectra were taken for two orthogonal polar-izations: parallel and perpendicular to the direction ofmechanical stress. In the following discussion, we willrefer to the horizontal and vertical directions inFigure 1c as x and y directions, respectively. Theoriginal heptamer structure is isotropic and thusshould exhibit identical spectra for the two polariza-tions. However, due to the slight shrinkage of theactual fabricated heptamer along the y direction, therewas a slight shift in spectra for the two orthogonalpolarizations. As shown in Figure 2, the position of theFano resonance, which presented itself as a dip in theextinction spectrum, was 829 nm for the x polarization,while the y polarization showed a dip at 838 nm. Toconfirm the small split was due to the imperfectfabrication, which resulted in a 3.7% shrinkage in they direction, numerical simulations using the commer-cial software COMSOLwere carried out. Figure 2 showsthat the simulations accounting for the vertical shrink-age in the actual fabricated structure precisely repro-duced the Fano resonance positions of experimentallymeasured spectra. The slight broadening in the experi-mental spectra was attributed to the slight size varia-tions (less than 2%) among the heptamers in the array.Also, the nanostructured gold is expected to exhibit

Figure 1. (a) Schematic diagram showing arrays of goldheptamers embedded in a PDMS membrane and beingstretched, (b) SEM image showing arrays of gold heptamersat low magnification, (c) close-up SEM image of single goldheptamer from the top view, and (d) SEM image of 45� tilted3 � 3 array of gold heptamers.

ARTIC

LE

CUI ET AL . VOL. 6 ’ NO. 3 ’ 2385–2393 ’ 2012

www.acsnano.org

2387

higher loss than the bulk, whose dielectric constantswere used in the simulations.35 This should account forsome of the discrepancies between the experimentaland simulation spectra. Overall, the agreement be-tween the simulation and experiment was excellent.The mechanical stress lowers the symmetry and

increases the degeneracy, producing distinct beha-viors for the two mutually orthogonal polarizations.Figure 2a shows experimentally measured and simu-lated extinction spectra at the induced mechanicalstrain values of 0%, 7%, 18%, and 30%with polarizationparallel to the mechanical stress (x direction), as in-dicated in the inset of Figure 2a. The Fano resonanceexhibited a moderate blue shift from 829 to 814 nm.Here, the mechanical strain is defined as the percentchange in the center-to-center spacing between ad-jacent gold cylinders along the x direction. Experimen-tally, it is not possible to directly measure the strains.We therefore determine the strain values indirectly byfirst measuring the changes in the length of theheptamer array, which could be directly imaged byoptical microscopy. In our experiments, the nominalvalues of strains measured this way were 11%, 28%,and 44%. However, the macroscopic geometrical de-formation of the PDMS may not translate perfectly tothe microscopic geometry of the heptamer structure.To determine the actual strain in the heptamers, weconducted mechanical simulations using COMSOL inwhich we modeled a single heptamer in a 900 nm �900 nm size computational cell and applied the

nominal strain values measured by the changes in arraysize. In the simulations, Young's modulus and Poisson'sratio were set to be 79 GPa and 0.44 for gold36,37 and360�870 kPa and 0.5 for PDMS.38 Further details on themechanical simulations may be found in the SupportingInformation. For the nominal strain values of 11%, 28%,and 44% measured by the optical microscope, the simu-lations resulted in actual strains of 7%, 18%, and 30%,respectively, and these values were then used for opticalsimulations. Figure 2b shows the experiment and simula-tion spectra for the polarization perpendicular to themechanical stress. In this case, the Fano resonance red-shifted significantly from838 to 874nm. In addition to thespectral shifts in the opposite directions, the extinctionspectra for the xpolarization showedanadditional featureat a shorter wavelength when the heptamer is undermechanical stress. This feature is noticeable at all strainvalues in simulations and also visible in experimentalspectra for strain values of 18% and 30%. In contrast, thisfeature is clearly absent in all simulated and experimentalspectra for ypolarizations. The samples havebeen subjectto multiple cycles of stress and release, and the opticalspectra remained unchanged. The data are available inthe Supporting Information.To gain insight into the observed behavior, we

conducted extensive theoretical studies by employingtwo complementary approaches. First, the eigen-modes of the heptamer structure were obtainedin the static limit by solving the boundary integraleigenvalue equations for the charge distribution.39,40

Figure 2. Experimentally measured and simulated extinction spectra for mechanical stress along the horizontal direction.Stretching direction and polarizations are shown in the insets. Solid lines show experiment results, and dashed lines showsimulation results.

ARTIC

LE

CUI ET AL . VOL. 6 ’ NO. 3 ’ 2385–2393 ’ 2012

www.acsnano.org

2388

This study allowedus to analyze the eigenmodeswith thegroup representation theory and thus correlate the sym-metry of the system with that of the eigenmodes. How-ever, since this approach is valid only in the static limit, itcannot properly describe the retardation effects thatmust be taken into account in a system with large sizefeatures such as the heptamers investigated in this paper.For this purpose, we also conducted numerical simula-tions using generalized multiparticle Mie theory.41 Sincethe multiparticle Mie theory can handle only sphericalparticles, the following theoretical studies deal with goldheptamersmadeof seven spheres, which should share allsalient featuresof thegold cylinder heptamers in Figure 2.In all simulations, the experimentally determined dielec-tric function of gold was used.35

To briefly describe the boundary integral approach,for a system composed of an arbitrary array of N

nanoparticles, the surface charge distribution σ j( rF)of its jth surface plasmonmode satisfies the boundary-integral eigenvalue equation:39,40

σj( rF) ¼ λj

2π

Iσ j( rFq)

rF � rFq

j rF � rFqj3 3n̂dSq (1)

where the integral is over the whole surfaces of the N

nanoparticles and n̂ is the unit vector normal to thesurfaces of the N particles at point rF. The eigenvalue λj

determines the system resonance frequency ωj throughthe electric permittivity ε of the array by the relationship

Re ε(ωj) ¼ εb1þ λj

1 � λj

!(2)

where εb is the electric permittivity of the mediumsurrounding the nanoparticles. The eigenfunction σ j( rF)describes the self-sustained surface charge distribution ofthe jth mode of the system. By solving eigenvalue eq 1numerically,39we can analyze the symmetryof themodes.Mirin et al. reported that the silver heptamer struc-

ture exhibits a Fano resonance as a result of theinterference between two modes, of which one issubradiant (or dark) and the other is superradiant (orbright).8 The same behavior is expected in the goldheptamer except the frequencies would be shifted dueto the difference in dielectric function between silverand gold. In the following presentation of our theor-etical analysis, we use the nomenclature based on thegroup representation theory.42 The heptamer structurediscussed in this paper has the symmetry of pointgroup D6h. Consequently, the eigenmodes can beindexed by the irreducible representations of D6h.Assuming the modes are excited by normally incidentlight with definite in-plane polarization, only the opti-cally active in-planemodes are considered in this work.Among the irreducible representations ofD6h, E1u is theonly one with a net dipole moment and thus opticallyactive. Furthermore, we consider only the two lowestenergy E1u modes in the unstressed gold heptamer

structure because all higher modes are masked off bystrong absorption by gold. In the first column ofFigure 3, we show the charge distribution of the twolowest energy E1u modes in an unstressed gold hepta-mer structure composed of seven identical goldspheres where the sphere diameter is 150 nm andgap between the spheres is 25 nm. Here the modesshown in Figure 3a,b belong to the lowest energy E1umode, and (d) and (e) to the second lowest E1u mode.Note that the E1u irreducible representation is a two-dimensional representation, and thus the E1u modesare doubly degenerate with two orthogonal stateshaving a net dipole moment in the x and y directions,respectively. Accordingly, the four charge distributionsshown in the first column of Figure 3 possess a netdipole moment where (a) and (d) are x-dipoles and (b)and (e) are y-dipoles. Depending on the energy and therelative alignment of the dipole moment of the centersphere to those of the six satellite spheres, the E1umodes can be classified as dark or bright modes. Thelower energy E1u mode shown in Figure 3a,b is a darkmode where the dipole moment of the center particlealigns against the dipolemoments of satellite particles,making the total dipole moment small. On the otherhand, the higher energy E1umode shown in Figure 3d,eis a bright mode where the dipole moments align

Figure 3. Evolution of charge distribution with uniaxialstress along the x direction. The first column shows thetwo lowest E1u modes of the unstressed heptamer and theB1u mode. The second and third columns show B2u and B3umodes of the heptamer under 10% and 30% strains, respec-tively. The resonance wavelengths of the heptamers arealso shown.

ARTIC

LE

CUI ET AL . VOL. 6 ’ NO. 3 ’ 2385–2393 ’ 2012

www.acsnano.org

2389

together to add up. The energies of these two E1umodes were found to be 2.394 and 2.459 eV or 517.8and 504.0 nm, respectively. These mode energy valueswould be accurate only for heptamers made of verysmall nanoparticles, as our model is valid in the staticlimit only. For larger sizes, retardation effects will shiftand broaden the modes. The resultant overlap andinterference between the two modes lead to a Fanoresonance. The bright mode will broaden much moresignificantly than the dark mode, resulting in a Fanoresonance that manifests itself in the form of a dip inthe extinction spectrum, as observed in Figure 2.When the heptamer is under uniaxial mechanical

stress, the symmetry of the system is lowered to D2h.The doubly degenerate E1u mode splits into two non-degenerate modes belonging to B2u and B3u irreduci-ble representations of the point group D2h. Figure 3shows the evolution of charge distribution as themechanical stress is applied along the x direction. Itclearly shows the original doubly degenerate modessplit into x-dipole (B3u) and y-dipole (B2u) modes.Remarkably, the nature of the modes is mostly pre-served. That is, the bright E1u mode splits into brightB2u and B3u modes, while the dark E1u mode spawnsdark B2u and B3u modes. Also, all modes shift to shorterwavelengths with increasing mechanical strain values.However, the B3u modes, which have dipole momentalong the direction of mechanical stress, shift morethan the B2u modes, with dipole moment perpendicu-lar to the mechanical stress. This leads to the polariza-tion dependence, as the B3u modes interact withx-polarized light and B2u with y-polarized light. There-fore, as the heptamer is stretched along the x direction,x-polarized light would show resonance features in theshorter wavelengths than the y-polarized light. Evenwhen the retardation effects are included and theresonance peaks broaden and shift, this general beha-vior survives and leads to the experimental observa-tion in Figure 2: the dip in the extinction spectrum dueto the Fano resonance blue-shifts for polarizationparallel to the direction of mechanical stress but red-shifts for polarization perpendicular to it.To confirm this, scattering and absorption cross

sections were calculated for heptamers made of spher-ical gold particles with diameters ranging from 50 to150 nm, using the generalizedmultiparticleMie theory.For small sphere sizes, both the absorption and scat-tering spectra exhibit a single peak centered around520 nm, which is a superposition of the two closelyspaced E1u modes. As the sphere sizes are increased,the two modes exhibit red shifts and broadening. Thebright mode experiences strong radiation damping,and therefore it quickly disappears in the absorptionspectra and produces a strong and broad band in thescattering spectra. The dark mode suffers from muchless radiation damping and thus persists as a peak inthe absorption spectra with relatively narrow line

width. In the scattering spectra, the overlap betweenthe two modes results in a dip, which is one of thewell-known signatures of Fano resonances. For spherediameters of 150 nm, both the absorption peak and thescattering dip occur at 650 nm.When the heptamers are subject to uniaxial mecha-

nical stress, the qualitative behaviors of the absorptionand scattering spectra remain similar. However, thepeak shift with increasing sphere diameters dependson the strain and polairzation. Figure 4 shows thescattering spectra for heptamers made of 150 nmgold spheres. With increasing mechanical stress alongthe x direction, the Fano resonance blue-shifts forx polarization but red-shifts for y polarization. The originof this distinct polarization dependence can be tracedback to the behaviors of B2u and B3u modes discussedearlier. The y-polarized light excites B2u modes, whichexhibit a much smaller blue shift than the B3u modes,which are excited by the x-polarized light. The retarda-tion effect in the large size heptamers then leads to redshifts, resulting in the B2u mode shifting to the red ofthe original E1u mode of the unstressed heptamerwhile the B3u mode appears to the blue of the E1umode. Also, the dip remains distinct in the y-polarizationspectra as the heptamer is stressed, whereas the dipbecomes shallower with stretching in the x-polarizationspectra. This behavior can also be seen from Figure 3a,b.The three modes shown in Figure 3b have almost

Figure 4. Scattering spectra of gold heptamers made of150 nm spheres for (a) x-polarization and (b) y-polarizationfor various mechanical strain.

ARTIC

LE

CUI ET AL . VOL. 6 ’ NO. 3 ’ 2385–2393 ’ 2012

www.acsnano.org

2390

identical charge distributions, and this shows that the B2umode originating from the dark E1u mode of the un-stressedheptamer remains dark undermechanical stress.From this observation, we anticipate the Fano dip wouldremain distinct under mechnical stress for y polarization.In contrast, Figure 3a shows that the dark B3u modebecomes brighter with stretching. Under mechanicalstress along the x direction, the coupling between thethree spheres along the x axis becomes significantlyweaker. This leads to the dipole moments of the twospheres on either side of the center sphere almostvanishing and also the dipole moments of the upperand lower spheres aligning almost vertically. As a result,the x-dipole of the center sphere is no longer canceled bythe surrounding spheres, and this consequently results ina diminishing dip in the scattering spectrum.In addition to the diminishing Fano dip, the scatter-

ing spectra for x polarization also show an additionaldip at shorter wavelengths. The additional dip isapparent in the spectra for 30% and 45% strain forx polarization but is clearly missing in all spectra fory polarization. This is a unique feature of heptamersunder stress that has not been reported before.The origin of this second dip can be found by thegroup theoretical analysis. Briefly reiterating, in theunstressed heptamer structure possessingD6h symme-try, only the E1u modes have nonzero dipole momentsand thus are optically active. When the symmetry islowered to D2h by uniaxial stress, the E1u modes splitinto B2u and B3u modes, which interact with x- andy-polarized light, respectively. Since the unstressedheptamer has two E1u modes in the frequency rangeof interest, we obtain two B2u and two B3u modes,producing a Fano resonance just as in the originalunstressed heptamer. However, what is missing in thisnarrative is that the optically inactive B1u mode in theunstressed heptamer becomes an optically active B3umode under uniaxial stress along the x direction. Asshown in Figure 3c, the charge distribution calculatedby the boundary integral method reveals that this B3umode is also a dark mode where the dipole moment ofthe center sphere aligns antiparallel against those ofthe satellite spheres, thereby producing a second Fanodip in the scattering spectra. In constrast, there are noother modes of the unstressed heptamer evolving intothe B2u mode within the frequency range we investi-gated, and thus we do not see any additional dip forthe y polarization. In Figure 4a, the second dip in thex-polarized spectra is clearly observed only at largestrain values. A possible reason the second dip is notresolved for small strain values is that the resonancewavelengths between the two B3u modes arising fromthe B1u and dark E1umodes are too close to be resolvedin the spectrum. The gold cylinder heptamer structurein Figure 2 showed the same general behavior, but inthis system the second Fano resonance was presenteven for small strain values.

To gain further insight, we calculated the modewavelengths of E1u and B1u modes of the unstressedheptamer by using the boundary integral method andfollowed how the wavelengths shift under mechanicalstress. As shown in Figure 5, all modes exhibit blueshifts. For y polarization, the two E1umodes simply shiftto shorter wavelengths, as shown in Figure 5b. Forx polarization, however, the behavior is complicated bythe presence of a B1u mode located close to the higherenergy E1u mode. As a result, while the low-energy E1umode remains largely isolated and shows a steady blueshift with increasing mechanical stress, the high-energy E1umode and B1umode, both of which becomeB3u modes under stress, interact strongly and exhibitanticrossing behavior. As shown in Figure 5a, the B3umode arising from the B1u mode stays at longerwavelengths than the B3umode spawned by the brightE1u mode until the mechanical strain value exceeds 15%,but the order is reversed at higher strain values. Carefulexamination of charge distributions near the anticrossingpoint revealed that the charge distributions showclear signs of mixing between the two modes. As shown

Figure 5. Resonant wavelengths of the three lowest energyoptically active modes as a function of mechanical strainvalues. Only the modes interacting with x polarization areshown in (a), and the modes interacting with y polarizationin (b). The resonance wavelengths are calculated for theheptamers made of 150 nm gold spheres by the boundaryvalue integral method.

ARTIC

LE

CUI ET AL . VOL. 6 ’ NO. 3 ’ 2385–2393 ’ 2012

www.acsnano.org

2391

in Figure 3c,d, when the heptamer is unstrained (0%), theB1u and E1u modes have distinct charge distributions.Among other things, the center particle of the E1u modearising from the B1u mode has a hexapole-like chargedistribution, while the center particle of the bright E1umode has a dipole-like charge distribution. For the mech-anical strain values of 10% at which anticrossing occurs,the center particle of the two B3u modes exhibits ahexapole-like charge distribution, clearly showing themixing of the two modes. For a larger strain value of30%,whichoccursbeyondanticrossing, the center particleof the lower B3u mode in Figure 3c recovers the purelydipole-like charge distribution, whereas the higher B3umode in Figure 3d has a hexapole-like distribution. Chargedistributions for additional strain valuesareprovided in theSupporting Information, as further evidence of modemixing. Additionally, the anticrossing behavior results inthe middle B3u mode wavelength becoming almost in-dependent of applied stress at high strain values,while thehighest energy B3u mode continues to show significantblue shift with increasing mechanical stress. This result,combined with the retardation effect, explains why thelonger wavelength Fano dip stops shifting to the blue butactually shows a slight red shift in Figure 4, whereas theshorter wavelength dip continues to show a blue shift.Anticrossing of mutually interacting states is a univer-

sal phenomenon. In quantum physics, anticrossing of aninteracting two-level system is well known.43 Anticross-ing is also routinely observed in optics. For example,anticrossing of a band-gap-guided mode and an index-guided mode in a photonic crystal waveguide is knownto produce slow group velocity.44 In plasmonics, anti-crossing behavior due to the strong mixing between amolecular exciton and a surface plasmon polariton hasbeen observed.45 In our structure, anticrossing is ob-served as the interaction between the two B1u modes istuned by the applied mechanical stress. It is intriguing todiscover that the originally optically inactive mode couldbecome optically active under mechanical stress andstrongly interact with the other optically active modes.This opens new possibilities for dynamically tuning andengineering resonances in theplasmonic nanostructures.

CONCLUSIONS

In conclusion, we have experimentally demon-strated dynamic tuning and symmetry lowering of

the Fano resonance in a gold heptamer structure andpresented an extensive theoretical study explainingthe observed behaviors. The original unstressed hep-tamer structure exhibits a distinct dip in the opticalextinction spectra due to the Fano resonance originat-ing from the two interacting E1u modes. When uniaxialstress is applied, the Fano dip blue-shifts for x polariza-tion but red-shifts for y polarization. Furthermore, thereappears a second dip for the x polarization when themechanical stress is large. Group theoretical analysisshowed that the uniaxial mechanical stress lowers theoriginal D6h symmetry of the heptamer to D2h, andconsequently the optically active doubly degenerateE1u mode of the unstressed heptamer splits into B3uand B2u modes, which have x- and y-direction dipolemoments and thus interact with x- and y-polarizedlight, respectively. Furthermore, the originally opticallyinactive B1u mode of the unstressed heptamer be-comes an optically active B3u mode under mechanicalstress and begins interacting stronglywith a nearby B3umode, strongly influencing the optical spectra ob-served under mechanical stress. The charge distribu-tion calculated by the boundary integral method in thestatic limit clearly showed how each mode shiftsand interacts with other modes. Accounting for theretardation effects for larger sizes by using the multi-particle Mie theory, the experimentally observed be-haviors were explained well and were also directlyconfirmed by the simulations using the finite elementmethod.Mechanical tuning of plasmonic nanostructures offers

anewpathway toactively tunablenanophotonicdevices.Mechanical tuning is particularly effective for plasmonicnanostructures exhibiting coupled plasmon resonance,which is highly sensitive to the structural parameters.Mechanical tuning also allows one to control the sym-metry of a nanostructure, upon which optical propertiesare critically dependent. It is intriguing to discover thatthe originally optically inactive mode could becomeoptically active under mechanical stress and stronglyinteract with the other optically activemodes. This opensnew possibilities for dynamically tuning and engineeringresonances in plasmonic nanostructures. Dynamic tun-ability and on-demand control afforded by mechanicallytunable nanophotonic devices could enable a new classof novel photonic devices.

METHODSThe heptamers were fabricated by an electron-beam litho-

graphy and lift-off process. The fabrication procedure startedwith a silicon wafer with a 100 nm Cr film thermally evaporatedon it. The Cr layer serves as a protection layer for the PDMSmembrane during the silicon dry etch process, preventingoveretching and reducing undesirable cracks in the PDMSmembrane. A bilayer of a polymethyl methacrylate (PMMA)resist and copolymer was coated on top of the Cr, and electron

beam lithography was carried out to define the heptamerstructure. Gold (80 nm) was then thermally evaporated ontothe patterned resist, and the final gold heptamer structure wasobtained by lift-off. A monolayer of (3-mercaptopropyl-)trimethoxysilane (MPTS) was deposited on the gold surfaceto function as an adhesion promotion layer and to improvecontact between gold and PDMS.18 Monolayer coating wascarried out in a 1% solution of MPTS in methanol. After over-night immersion in the solution, the gold heptamer was rinsed

ARTIC

LE

CUI ET AL . VOL. 6 ’ NO. 3 ’ 2385–2393 ’ 2012

www.acsnano.org

2392

inmethanol to remove excessMPTS on the surface. PDMSwith athickness of 850 μmwas spin-coated on the heptamer structureto completely cover the heptamer structure and was cured at100 �C for 2 h. Finally, the silicon substrate was completely dryetched by SF6-based reactive ion etching, leaving the heptamer-embedded PDMS membrane with Cr protection layer atthe bottom. After removal of 100 nm Cr, a completely free-standing PDMS membrane with a heptamer structure wasobtained.Mechanical stress was applied to the PDMS membrane by a

custom-made device, where an aperture at the center issurrounded by a rigid frame and a linear translation stagecontrolled by a micrometer. The length, width, and height ofthe platform are 7.5, 5.0, and 1.3 cm, respectively. The aperturein the center over which the flexible heptamer structure is to beattached has a square shape with a side of 2.5 cm. One end ofthe PDMS membrane is glued to the rigid frame, and the otherend to the movable translation stage with which mechanicalstress is applied in a controlled fashion.

Conflict of Interest: The authors declare no competingfinancial interest.

Acknowledgment. This work was supported by NationalScience Foundation (BES 0608934).

Supporting Information Available: Additional informationon the mechanical simulation, repeatability experiment, andcharge distribution for additional strain value is available. Thismaterial is available free of charge via the Internet at http://pubs.acs.org.

REFERENCES AND NOTES1. Pakizeh, T.; Kall, M. Unidirectional Ultracompact Optical

Nanoantennas. Nano Lett. 2009, 9, 2343–2349.2. Brown, L. V.; Sobhani, H.; Lassiter, J. B.; Norlander, P.; Halas,

N. J. Heterodimers: Plasmonic Properties of MismatchedNanoparticle Pairs. ACS Nano 2010, 4, 819–832.

3. Mastroianni, A. J.; Claridge, S. A.; Alivisatos, A. P. Pyramidaland Chiral Groupings of Gold Nanocrystals AssembledUsing DNA Scaffolds. J. Am. Chem. Soc. 2009, 131, 8455–8459.

4. Zhang, Z.; Weber-Bargioni, A.; Wu, S. W.; Dhuey, S.; Cabrini,S.; Schuck, P. J. Manipulating Nanoscale Light Fields withAsymmetric Bowtie Nano-Colorsorter. Nano Lett. 2009, 9,4505–4509.

5. Britt Lassiter, J.; Sobhani, H.; Fan, J. A.; Kundu, J.; Capasso, F.;Nordlander, P.; Halas, N. J. Fano Resonance in PlasmonicNanoclusters: Geometrical and Chemical Tunability. NanoLett. 2010, 10, 3184–3189.

6. Hentschel, M.; Saliba, M.; Vogelgesang, R.; Giessen, H.; PaulAlivisatos, A.; Liu,N. Transition from Isolated toCollectiveModesin Plasmonic Oligomers. Nano Lett. 2010, 10, 2721–2726.

7. Fan, J. A.; Wu, C. H.; Bao, K.; Bao, J. M.; Bardhan, R.; Halas,N. J.; Manoharan, V. N.; Nordlander, P.; Shvets, G.; Cappasso,F. Self-Assembled Plasmonic Nanoparticle Clusters. Science2010, 328, 1135–1138.

8. Mirin, N. A.; Bao, K; Nordlander, P. Fano Resonances inPlasmonic Nanoparticle Aggregates. J. Phys. Chem. A2009, 113, 4028–4034.

9. Alonso-Gonzalez, P.; Schnell, M.; Sarriugarte, P.; Sobhani,H.; Wu, C.; Arju, N.; Khanikaev, A.; Golmar, F.; Albella, P.;Arzubiaga, L.; et al. Real-Space Mapping of Fano Interfer-ence in Plasmonic Metamolecules. Nano Lett. 2011, 11,3922–3926.

10. Ou, F. S.; Hu, M.; Naumov, I.; Kim, A.; Wu, W.; Bratkovsky,A. M.; Li, X.; Williams, R. S.; Li, Z. Hot-Spot Engineering inPolygonal Nanofinger Assemblies for Surface EnhancedRaman Spectroscopy. Nano Lett. 2011, 11, 2538–2542.

11. Bao, K.; Mirin, N. A.; Nordlander, P. Fano Resonances inPlanar Silver Nanosphere Clusters. Appl. Phys. A: Mater. Sci.Process. 2010, 100, 333–339.

12. Gallinet, B.; Martin, O. J. F. Influence of ElectromagneticInteractions on the Line Shape of Plasmonic Fano Reso-nances. ACS Nano 2011, 5, 8999–9008.

13. Dregely, D.; Hentschel, M.; Giessen, H. Excitation andTuning of Higher-Order Fano Resonances in PlasmonicOligomer Clusters. ACS Nano 2011, 5, 8202–8211.

14. Yan, B.; Boriskina, S. V.; Reinhard, B. M. Optimizing GoldNanoparticle Cluster Configurations (n e 7) for ArrayApplications. J. Phys. Chem. C 2011, 115, 4578–4583.

15. Park, W.; Lee, J.-B. Mechanically Tunable Photonic CrystalStructure. Appl. Phys. Lett. 2004, 85, 4845–4847.

16. Wong, C. W.; Rakich, P. T.; Johnson, S. G.; Qi, M.; Smith, H. I.;Ippen, E. P.; Kimerling, L. C.; Jeon, Y.; Barbastathis, G.; Kim,S.-G. Strain-Tunable Photonic Band Gap Microcavities inOptical Waveguides. Appl. Phys. Lett. 2004, 84, 1242–1244.

17. Cui, Y.; Tamma, V. A.; Lee, J.-B.; Park,W.Mechanically TunableNegative-Index Photonic Crystal Lens. IEEE Photonics J.2010, 2, 1003–1012.

18. Pryce, I. M.; Aydin, K.; Kelaita, Y. A.; Briggs, R. M.; Atwater, H. A.Highly Strained Compliant Optical Metamaterials with LargeFrequency Tunability. Nano Lett. 2010, 10, 4222–4227.

19. Ou, J. Y.; Plum, E.; Jiang, L.; Zheludev, N. I. ReconfigurablePhotonic Metamaterials. Nano Lett. 2011, 11, 2142–2144.

20. Huang, F.; Baumberg, J. J. Actively Tuned Plasmons onElastomerically Driven Au Nanoparticle Dimers. Nano Lett.2010, 10, 1787–1792.

21. Fano, U. Effects of Configuration Interaction on Intensitiesand Phase Shifts. Phys. Rev. 1961, 124, 1866–1878.

22. Bulka, B. R.; Stefanski, P. Fano and Kondo Resonance inElectronic Current through Nanodevices. Phys. Rev. Lett.2001, 86, 5128–5131.

23. Kobayashi, K.; Aikawa, H.; Sano, A.; Katsumoto, S.; Iye, Y.Fano Resonance in a Quantum Wire with a Side-CoupledQuantum Dot. Phys. Rev. B 2004, 70, 035319.

24. Ladron de Guevara, M. L.; Claro, F.; Orellana, P. A. GhostFano Resonance in a Double Quantum Dot MoleculeAttached to Leads. Phys. Rev. B 2003, 67, 195335.

25. Hao, F.; Sonnefraud, Y.; Van Dorpe, P.; Maier, S. A.; Halas,N. J.; Nordlander, P. Symmetry Breaking in PlasmonicNanocavities: Subradiant LSPR Sensing and a TunableFano Resonance. Nano Lett. 2008, 8, 3983–3988.

26. Hao, F.; Nordlander, P.; Sonnefraud, Y.; Van Dorpe, P.;Mailer, S. A. Tunability of Subradiant Dipolar and Fano-Type Plasmon Resonance in Metallic Ring/Disk Cavities:Implications for Nanoscale Optical Sensing. ACS Nano2009, 3, 643–652.

27. Verellen, N.; Sonnefraud, Y.; Sobhani, H.; Hao, F.;Moshchalkov,V. V.; Van Dorpe, P.; Nordlander, P.; Maier, S. A. Fano Reso-nances in Individual Coherent Plasmonic Nanocavities. NanoLett. 2009, 9, 1663–1667.

28. Liu, N.; Langguth, L.; Weiss, T.; Kastel, J.; Fleischhauer, M.;Pfau, T.; Giessen, H. Plasmonic Analogue of Electromagnet-ically Induced Transparency at the Drude Damping Limit.Nat. Mater. 2009, 8, 758–762.

29. Christ, A.; Ekinci, Y.; Solak, H. H.; Gippius, N. A.; Tikhodeev,S. G.; Martin, O. J. F. Controlling the Fano Interference in aPlasmonic Lattice. Phys. Rev. B 2007, 76, 201405(R).

30. Lukyanchuk, B.; Zheludev, N. I.; Maier, S. A.; Halas, N. J.;Norlander, P.; Giessen, H.; Chong, C. T. The Fano Resonancein a Plasmonic Nanostructures and Metamaterials. Nat.Mater. 2010, 9, 707–715.

31. Bachelier, G.; Russier-Antoine, I.; Benichou, E.; Jonin, C.; DelFatti, N.; Vallee, F.; Brevet, P.-F. Fano Profiles Induced byNear-Field Coupling in HeterogeneousDimers of Gold andSilver Nanoparticles. Phys. Rev. Lett. 2008, 101, 197401.

32. Tetz, K. A.; Pang, L.; Fainman, Y. High-Resolution SurfacePlasmon Resonance Sensor Based on Linewidth-OptimizedNanoholeArrayTransmittance.Opt. Expr.2006,31, 1528–1530.

33. Chuntonov, L.; Haran, G. Trimeric Plasmonic Molecules:The Role of Symmetry. Nano Lett. 2011, 11, 2440–2445.

34. Chuntonov, L.; Haran, G. Effect of Symmetry Breakingon the Mode Structure of Trimeric Plasmonic Molecules.J. Phys. Chem. C 2011, 115, 19488–19495.

35. Johnson, P. B.; Christy, R. W. Optical Constants of the NobleMetals. Phys. Rev. B 1972, 15, 4370.

36. Volker, C. A.; Lilleodden, E. T. Size Effects in the Deforma-tion of Sub-Micron Au Columns. Philos. Mag. 2006, 86,5567–5579.

ARTIC

LE

CUI ET AL . VOL. 6 ’ NO. 3 ’ 2385–2393 ’ 2012

www.acsnano.org

2393

37. Robinson, R. G. R.; Stephens, K. G.; Wilson, I. H. The Effect ofIon Bombardment on the Strain Gauge Factor of Thin GoldFilms. Thin Solid Films 1975, 27, 251–262.

38. Armani, D.; Liu, C.; Aluru, N. Re-Configurable Fluid Circuitsby PDMS Elastomer Micromachining. Proc. IEEE MEMS991999, 222–227.

39. Mayergoyz, I. D.; Fredkin, D. R.; Zhang, Z. Electrostatic(Plasmon) Resonances in Nanoparticles. Phys. Rev. B2005, 72, 155412.

40. Gomez, D. E.; Venron, K. C.; Davis, T. J. Symmetry Effects onthe Optical Coupling between Plasmonic Nanoparticleswith Applications to Hierarchical Structures. Phys. Rev. B2010, 81, 075414.

41. Yu, Y. I.; Gustafson, B. A. S. A Generalized MultiparticleMie-Solution: Further Experimental Verification. J. QuantumSpectrosc. RA. 2001, 70, 395.

42. Tinkham Group Theory and QuantumMechanics; McGraw-Hill, 1964.

43. Feynman, R. P.; Leighton, R. B.; Sands, M. In The FeynmanLectures, Vol. 3. Quantum Mechanics; Addison-WesleyPublishing Co.: London, 1970; Chapters 9�11.

44. Notomi, M.; Yamada, K.; Shinya, A.; Takahashi, J.; Takahashi,C.; Yokohama, I. Extremely Large Group-Velocity Disper-sion of Line-Defect Waveguides in Photonic Crystal Slabs.Phys. Rev. Lett. 2001, 87, 253902.

45. Dintinger, J.; Klein, S.; Bustos, F.; Barnes, W. L.; Ebbesen,T. W. Strong Coupling between Surface Plasmon-Polaritons and Organic Molecules in SubwavelengthHole Arrays. Phys. Rev. B 2005, 71, 035424.

ARTIC

LE